Super-resolution reconstruction using kernel regression and feature-driven prior in a charge-coupled device sensor system

This paper proposes a novel super-resolution (SR) algorithm embedded in a CCD (charge-coupled device) sensor system to enhance imaging resolution. In this system, multiple CCD sensors acquire images simultaneously, complement each other the information, and avoid information shortage in a single sensor. The proposed SR algorithm adopts the trilateral kernel regression for interpolation, which allows for spatial distance, photometric difference, and confidence of pixels. Then a maximum a priori (MAP) optimization is employed for image restoration using feature-driven prior which completely depends on the statistical characteristics of the image itself, thus the reconstruction is more accurate. The visual effect and index of experimental results show the proposed algorithm is effective.